An expository paper featuring the science behind Petro.ai, Lithologically-Controlled Variations of the Least Principal Stress with Depth and Resultant Frac Fingerprints During Multi-Stage Hydraulic Fracturing, authored by Petro.ai and Stanford University’s Dr. Mark Zoback, will be featured at the upcoming URTeC Conference in Houston, Texas on June 20-22, 2022.
“The point of the paper,” Dr. Brendon Hall, VP of Geoscience and one of the paper’s authors, explains, “is making the case that a frac fingerprint is controlled by stress. Stress changes in the earth because of lithologic changes, because of changes in rock type meaning going from shale to sandstones and those changes really guide which areas are affected by hydraulic fracturing.
“The Frac Fingerprint Zoback defines as the pattern of hydraulic fracturing propagation. The drainage volumes that result from that are largely controlled by the variation of stress, in particular the minimum horizontal stress. The location of the stage with respect to that depth profile will be very important in determining what that complex pattern looks like. Petro.ai calls that complex pattern the Frac Fingerprint. It’s an expression of the pattern of fracture propagation in the gun barrel space, looking down the length of the lateral.
“The paper goes on to show evidence for this. Observational evidence, evidence from numerical simulation using ResFrac and then demonstrations from Petro.ai using the simulations we do to compare with production and decisions about optimization that you as an operator can make using the Frac Fingerprint.
“The drainage area is a really important feature for predicting the production of the wells. This makes sense when you think about what’s actually going on. The amount of hydrocarbon molecules you’re going to extract from the ground is going to be proportional to the amount of surface area you create within the earth during the hydraulic fracturing stimulation operation.
“You create these hydraulic fractures. They go out and connect with the natural fracture network and create a surface network of these cracks and fractures. Over time, hydrocarbon molecules can diffuse out of the rock formation, a short distance over their lifetime, into this fracture network and actually find their way back to the wellbore.
“The more drainage area, the more stimulated region you connect with, the more you’re going to be able to produce. That’s hard to observe directly. One important component in estimating the drainage area is knowing the overall fingerprint of the hydraulic fracturing operation.”
“I would also emphasize,” Kyle LaMotta, VP of Analytics adds, “the feature that we generate in Petro.ai called the Frac Fingerprint is an approximation of the volume that is available to be drained by a well. It’s different than a reservoir simulator or a frac simulator and it ends up being a very important feature in multivariate models. It considers lots of other variables that would otherwise have to be considered separately such as well spacing, well generation as whether it’s a parent or a child. It handles the complexity of the 4D nature, three dimensions plus time. We put that into one feature that ends up being highly predictive of the well’s performance.
“The Frac Fingerprint is vitally important to the industry. How is an operator estimating their productivity without using drainage? The most common way is to make a type curve. They would pick 15 or 20 wells in the area that they think represents the new wells that they’re going to drill. Then they take an average or a P50 of those other wells. And depending on whether it’s a child or a sibling they’ll apply degradation factors. They might take 15 wells in the area but some are going to be children. Now some might be parents so they’re going to apply a degradation factor like maybe cut off 20% if it’s directly offsetting the parent. If it's not maybe, they’ll take off 10%. It’s an aggregate of existing data and then an adjustment. It’s all scenario based, ‘we’ve seen in the data that child wells typically perform 30% worse than parent, so we’re going to scale our type curves by 30%. ’It’s not an exact science.